Longitudinal interpolation of parameters characterizing channel geometry by piece-wise polynomial and universal kriging methods: effect on flow modeling

Channel geometry often is described by a set of longitudinally varying parameters measured at a set of survey stations. To support flow modeling at arbitrary resolution, three methods of parameter interpolation are described including piece-wise linear interpolation, monotone piece-wise-cubic Hermitian interpolation, and universal kriging. The latter gives parameter estimates that minimize the mean square error of the interpolator, and therefore can be used as a standard against which the accuracy of polynomial methods can be assessed. Based on the application of these methods to a dataset describing cross-sectional properties at 283 stations, piece-wise linear interpolation gives parameter estimates that closely track universal kriging estimates and therefore this method is recommended for routine modeling purposes. Piece-wise-cubic interpolation gives parameter estimates that do not track as well. Differences between cubic and kriging estimates were found to be 2-10 times larger than differences between linear and kriging parameter estimates. In the context of one-dimensional flow modeling, the sensitivity of steady state water level predictions to the channel bed interpolator is comparable to a 5% change in the Manning coefficient. (C) 2004 Elsevier Ltd. All rights reserved.